STEM

该数据集名为STEM，主要用于评估神经模型和人类在各类与STEM相关任务上的表现，特别关注对视觉和语言信息的理解，这对于掌握STEM技能至关重要。此外，该数据集揭示了神经模型与人类学生在表现上的差距，突显了在复杂推理和抽象知识掌握方面的挑战。尽管具体规模未提及，但该数据集的规模较大。其任务是对使用神经模型的STEM技能进行评估，并与人类的表现进行比较。

This dataset, named STEM, is primarily designed to evaluate the performance of both neural models and human students across various STEM-related tasks, with a particular focus on the comprehension of visual and linguistic information—an ability critical for mastering STEM skills. Additionally, this dataset reveals the performance gap between neural models and human students, highlighting the challenges in complex reasoning and the mastery of abstract knowledge. Although its exact scale is not specified, this is a large-scale dataset. Its core task is to assess the performance of neural models on STEM-related tasks and compare such performance against that of human participants.